[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
|
Re: [rt-pmc] 08 Sep 2010 - RT PMC call follow-up
|
I had a talk with Barb about this a bit and wanted to follow up on
this thread with this blurb from our conversation...
my current frustration with the IP policy/procedure is the existence
of a 'safe on the build server' classification for some things. In my
little slice of heaven we write software that builds straight out of
svn...and the current definition of 'distribution' includes what is in
svn...which means that I am unable to make use of this 'safe on the
build server' classification of IP. From what I hear the whole 'safe
on the build server' exists because there is an explicit understanding
that the builds produced on the build server are not 'eclipse
distributions'.
so I don't understand how there exists a case where a build occurring
on the eclipse build server is not a distribution but somehow software
being built from something checked out from SVN is some how an
'eclipse distribution'. I understand that is absurd and its not
'really' but if I want to open a CQ to use something like apache
directory service its basically a rubber stamp if I have it on the
build server and use it there, but if my build system has an actual
'dependency' declared on it for usage in a test case (not going into
my distribution at _all_, just a unit test) it is suddenly a full
fledged dependency (because there is no distinguishing between test
scope or distribution scope).
cheers,
jesse
--
jesse mcconnell
jesse.mcconnell@xxxxxxxxx
On Mon, Sep 13, 2010 at 15:45, Jeff McAffer <jeff@xxxxxxxxxxxxxxxxx> wrote:
> Hey Barb,
> Thanks for the additional info. Comments embedded.
> Jeff
> On 2010-09-13, at 2:11 PM, Barb Cochrane wrote:
>
> 1) Eclipse-developed test code deposited in the repository calling
> third party test code which does not reside on Eclipse servers.
>
>
>
> Ø In the cases described, third party code could be characterized as a
> “workswith” or perhaps “exempt pre-req”
>
> Can you clarify "could"? Is it "should" or "must"?
> As I recall from the discussion there were various shades here. There are
> a) test setups that just need some other function to be present but do not
> reference this function directly in any code.
> b) test setups that need specific function to be present but still do not
> reference the function directly
> c) test setups that directly reference the third party libraries but that
> function is not part of the actual released output of the project
> Other scenarios where the third party references come from (to be) released
> project output are easier to handle as classical dependencies. The subtlety
> in the above comes from the absence/presence of direct references and the
> fact that test code is not generally released so is not typical consumable
> project output. That is, we don't expect any significant number of people to
> get this code (see case 2 below). Under that model, scenarios a-c above
> would be instances of case 2.
>
> Ø The process outlined in the Guidelines for the Review of Third Party
> Dependencies would apply (as there is no distinction made based on the type
> of code involved)
>
> Ø It is our understanding that the dynamic nature of calls to these
> third party packages would make the task of creating CQs for each package an
> extraordinarily heavy burden
>
> Ø The IP team is in the process of reviewing the IP Policy to try to
> identify what we can do to alleviate this burden. Janet will raise any
> suggested changes to the Policy to the IP Advisory Committee as necessary.
>
>
>
> 2) A risk was identified in that there are a small number of
> bleeding-edge developers (low single percentage of users) who may download
> code from the build server or from the Eclipse repository before it is
> “officially” released by a project. In either case, the downloader may
> consume code that has not been fully reviewed / approved by the IP team.
> We appreciate this risk being highlighted but feel the additional controls
> necessary to eliminate this small risk would not justify the adverse impact
> such additional controls would have on the community.
>
>
>
> 3) There was a question raised as to why CQs are required for
> “workswith” dependencies, but not their further dependencies. In the case
> of dependencies, we are trying to highlight to our downstream consumers the
> Eclipse Project dependencies in such a way so that they may investigate
> further and make decisions accordingly. In so doing, we must determine what
> level of information is reasonable to provide given our resource
> constraints. It was felt that the first level of information regarding
> dependencies struck a reasonable balance given our current staffing levels.
>
>
>
>
> I hope that helps. If an additional conference call would be useful,
> please let me know. I hope to update you further on the first topic in the
> next couple of weeks.
>
>
>
> Cheers,
>
>
>
> Barb Cochrane
>
> Eclipse Foundation, Inc.
>
> Phone 613.224.9461 ext 232
>
>
>
> Eclipse Summit Europe, Nov 2-4
>
> http://www.eclipsecon.org/summiteurope2010/
>
>
>
> _______________________________________________
> rt-pmc mailing list
> rt-pmc@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/rt-pmc
>
>
> _______________________________________________
> rt-pmc mailing list
> rt-pmc@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/rt-pmc
>
>
